Silk 丝路 silk

Silk is an AI multi-model gateway running on your desktop (Tauri 2 + Rust/Axum). All the hassles you face when managing multiple AI providers—incompatible protocols, scattered API keys, or an entire service crashing due to rate limiting—are solved with a single local HTTP endpoint: http://127.0.0.1:1877. ✨ Feature Highlights - 🌐 Unified Access — Your ChatBox, LobeChat, VS Code plugins, and any OpenAI-compatible clients... just point them all to one address. Silk automatically routes to the correct provider based on the model name. - 🔄 Automatic Protocol Conversion — Send requests in OpenAI Chat format, and even Claude will respond correctly. Seamlessly convert between OpenAI Chat, Claude Messages, and OpenAI Responses protocols without memorizing three sets of API docs. - 🛡️ Three-Level Failover — Upstream timeout → Auto-retry (exponential backoff) → Switch API Key → Switch Provider. Users remain unaware, and your AI experience stays uninterrupted. - 🔐 Local-First, Privacy-Centric — All data is stored in a local SQLite database, and API keys are encrypted with AES-GCM. With no cloud components, you have full control over every request. - 🔌 Pluggable Middleware — Gateway middleware loads on demand; turn off what you don't need (Token-saving extensions: coming soon). - 🖥️ Graphical Management — Out-of-the-box desktop UI (Vue 3 + NaiveUI). Manage providers, routing rules, and view logs without writing a single line of configuration. 🔄 Protocol Support Silk supports mutual conversion between three major protocols: OpenAI Chat (openai), Claude Messages (messages), and OpenAI Responses (responses). Requests entering via any protocol can be correctly forwarded to the upstream in the target format, with the response converted back to the client's protocol. | Protocol | Client Endpoint | Status | | ---------------------- | -------------------- | ----------- | | OpenAI Chat Completion | /v1/chat/completions | ✅ Supported | | Claude Messages | /v1/messages | ✅ Supported | | OpenAI Responses | /v1/responses | ✅ Supported | | Gemini | — | 🚧 In Dev | Mutual conversion between the three protocols has been verified through practical debugging with OpenCode + Agnes (including streaming SSE and thinking process events). Gemini protocol conversion is still under development and will be added in future versions.